Understand and optimize costs in Amazon Managed Service for Prometheus - Amazon Managed Service for Prometheus
What contributes to my costs?What is the best way to lower my costs? How do I lower ingestion costs?What is the best way to lower my query costs?If I decrease the retention period of my metrics, will that help reduce my total bill?How can I keep my alert query costs low?What metrics can I use to monitor my costs?Can I check my bill at any time?Why is my bill higher at the beginning of the month than at the end of the month?I deleted all my Amazon Managed Service for Prometheus workspaces, but I still seem to be getting charged. What might be happening?

Understand and optimize costs in Amazon Managed Service for Prometheus

The following frequently asked questions and their answers may be helpful in understanding and optimizing costs associated with Amazon Managed Service for Prometheus.

What contributes to my costs?

For most customers, metric ingestion contributes the majority of costs. Customers with high query usage will also see some cost based on query samples processed, with metrics storage being a small driver of overall costs. For more information about the prices for each of these, see Pricing in the Amazon Managed Service for Prometheus product page.

What is the best way to lower my costs? How do I lower ingestion costs?

Ingestion rates (not storage of the metrics) is the majority of costs for most customers. You can reduce ingestion rates by reducing the collection frequency (increasing the collection interval) or by reducing the number of active series ingested.

You can increase the collection (scraping) interval from your collection agent: Both the Prometheus server (running in Agent mode) and the AWS Distro for OpenTelemetry (ADOT) collector support the scrape_interval configuration. For example, increasing the collection interval from 30 seconds to 60 seconds will reduce your ingestion usage by half.

You can also filter the metrics sent to Amazon Managed Service for Prometheus by using the <relabel_config>. For more information about relabeling in the Prometheus agent configuration, see https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config in the Prometheus documentation.

What is the best way to lower my query costs?

Query costs are based on the number of samples processed. You can reduce the frequency of queries to reduce your query costs.

To get more visibility into the queries that are contributing the most to your query costs, you can reach out to file a ticket with your support contact. The Amazon Managed Service for Prometheus team can help you understand the queries that are contributing the most to your costs.

If I decrease the retention period of my metrics, will that help reduce my total bill?

You can reduce your retention period, however, this is unlikely to substantially reduce your costs.

If you want to reduce (or increase) your retention period, you can file a service limit request to change the Retention time for ingested data quota.

How can I keep my alert query costs low?

Alerting creates queries against your data, which add to your query costs. Here are some strategies that you can use to optimize your alert queries, and keep your costs lower.

  • Use Amazon Managed Service for Prometheus alerting – Alerting systems external to Amazon Managed Service for Prometheus may require additional queries to add resiliency or high availability, as the external service queries the metrics from multiple availability zones or regions. This includes alerting in Grafana for high availability. This can multiply your cost by three times or more. The alerting in Amazon Managed Service for Prometheus is optimized and will give you high availability and resiliency with the fewest number of queries.

    We recommend using the native alerting in Amazon Managed Service for Prometheus rather than external alerting systems.

  • Optimize your alert interval – One quick way to optimize your alert queries is to increase the auto-refresh interval. If you have an alert that queries every minute, but is only needed every five minutes, increasing the auto-refresh interval could save you five times your query costs for that alert.

  • Use an optimal lookback – A larger lookback window in your query increases the costs of the query, as it pulls more data. Ensure that the lookback window in your PromQL query is reasonably sized for the data you need to alert. For example, in the following rule, the expression includes a ten minute lookback window:

        - alert: metric:alerting_rule
      expr: avg(rate(container_cpu_usage_seconds_total[10m])) > 0
      for: 2m

    Changing the expr to avg(rate(container_cpu_usage_seconds_total[5m])) > 0 can help to reduce your query costs.

In general, look at your alerting rules and make sure that you are alerting on the best metrics for your service. It's easy to create overlapping alerts on the same metrics or multiple alerts that give you the same information, especially as you add alerts over time. If you find that you often see groups of alerts happening at the same time, it's possible that you can optimize your alerts and not include all of them.

These suggestions can help you to reduce costs. Ultimately, you must balance the costs with creating the right set of alerts for understanding the state of your system.

For more information about alerting in Amazon Managed Service for Prometheus, see Managing and forwarding alerts in Amazon Managed Service for Prometheus with alert manager.

What metrics can I use to monitor my costs?

Monitor IngestionRate in Amazon CloudWatch to track your ingestion costs. For more information about monitoring Amazon Managed Service for Prometheus metrics in CloudWatch, see Use CloudWatch metrics to monitor Amazon Managed Service for Prometheus resources.

Can I check my bill at any time?

The AWS Cost and Usage Report tracks your AWS usage and provides estimated charges associated with your account within a billing period. For more information, see What are AWS Cost and Usage Reports? in the AWS Cost and Usage Reports User Guide

Why is my bill higher at the beginning of the month than at the end of the month?

Amazon Managed Service for Prometheus has a tiered pricing model for ingestion, which results in costs in your initial usage being higher. As your usage reaches higher ingest tiers, with lower costs, your costs are lower. For more information about pricing, including ingest tiers, see Pricing in the Amazon Managed Service for Prometheus product page.

Note

  • Tiers are for usage within a region, not across regions. Usage within a region must reach the next tier to use the lower rate.

  • In an organization in AWS Organizations, tier usage is tallied per payer account, not per account (the payer account is always the organization management account). When the total ingested metrics (within a region) for all accounts in an organization reaches the next tier, all accounts are charged the lower rate.

I deleted all my Amazon Managed Service for Prometheus workspaces, but I still seem to be getting charged. What might be happening?

One possibility in this case is that you still have AWS managed scrapers that are setup to send metrics to your deleted workspaces. Follow the instructions to Find and delete scrapers.